Recent deep learning (DL) detectors adopted by radar or sonar (RS) are normally trained with transfer learning, where the typical workflow is to pretrain a convolutional neural network (CNN) on external large-scale classification datasets (e.g., ImageNet) as the backbone and then finetune the entire detector on detection datasets. Though transfer learning could effectively avoid overfitting, transferred models are usually redundant and might not generalize well on RS datasets. To achieve high generalization and to eliminate the dependence on transfer learning, a self-trained target detection method is established by including Automatic Deep Learning (AutoDL) to design optimal detectors. This self-trained target detection consists of three stages. First, a derived classification dataset (DCD) consisting of image blocks of targets and backgrounds is derived from detection datasets. Then, a memory-efficient Differentiable Architecture Search algorithm with flexible search space and large inputs (FL-DARTS), which is characterized by its predefined multistride convolutions, poolings, and unique super-structure, is proposed to automatically design and self-train optimal CNNs on DCDs. Finally, self-trained AutoDL detectors are implemented with the automatic backbone designed by FL-DARTS. We evaluated three self-trained AutoDL detectors on the public SAR ship detection dataset (SSDD) and the self-made sonar common target detection dataset (SCTD). The experiments show that while the number of parameters of automatic backbones designed for SSDD and SCTD are only 11.8% and 15.2% of that of ResNet50, self-trained AutoDL detectors implemented with automatic backbones significantly outperform their transfer learning detectors and achieve state-of-the-art detection precisions and high detection speeds. Data, codes are publicly available.

In multi-receiver Synthetic Aperture Sonars (SAS), the traditional non-stop-hop-stop model normally ignores the intra-pulse motion of sonar platform. However, such intra-pulse motion brings the Doppler scale effect to the echo signal, which will lead to the change of Carrier Frequency (CF) and Frequency Modulation (FM) rate of echo signal and may eventually result in image defocus. In this paper, considering the intra-pulse motion, the time-varying time delay and range history of multi-receiver SAS were derived by introducing the range time into the range history, and then the Doppler scale signal model was established. After approximating the derived time-varying range history, the two-dimensional (2D) spectrum was derived by using the shear theorem. When compared with the traditional 2D spectrum, an interesting phenomenon was found that the change of FM rate has no effect on 2D spectrum. Actually, only two additional terms, i.e., range cell migration (ARCM) and azimuth modulation (AAM), were introduced into the new 2D spectrum due to the change of CF. Besides, the condition of neglecting the ARCM was given by further analyzing the 2D spectrum, while the AAM is usually small and can be easily compensated together with the ARCM correction. Finally, a Chirp-z transform (CZT) algorithm for the Doppler scale signal model was proposed to correct the ARCM, which can simultaneously correct the ARCM and compensate the AAM (although not required) in 2D frequency domain by phase multiplication. The effectiveness of the signal model and proposed algorithm is verified by simulations and field data.

The interpolation-free chirp z-transform (CZT) algorithm is efficient but requires approximating the accurate two-dimensional (2-D) spectrum’s phase. However, this approximation is invalid for data from multiple-receiver synthetic aperture sonar (MRSAS) systems with wide beamwidths, large fractional bandwidths, and wide subswaths. To address the issue above, this paper derives a new range history and 2-D spectrum, and proposes a new method to reduce the approximation error of the 2-D spectrum’s phase. This paper first represents the range history as a combination of basic functions and derives a new 2-D spectrum without any approximation. After the receiver dependence of the 2-D spectrum’s phase is partially reduced, the echo data is divided into multiple subapertures in the receiver-domain to ensure that the 2-D spectrum’s phase difference of different receivers within each subaperture can be ignored. A frequency-domain filter is designed based on the 2-D spectrum of the reference receiver to compensate for all receivers’ data in the subaperture. Therefore, only one imaging process is needed in a subaperture, rather than receiver-by-receiver imaging. The 2-D spectrum’s phase is expanded up to the third range frequency and its error is limited by the subswath-subband processing rather than considering the higher-order terms in Taylor expansion. The quadratic and cubic terms in each subband are linearized through subband-wise linear approximation (SLA) to facilitate the CZT algorithm. Finally, the high-resolution image of the whole swath is obtained through the subimage fusion. The superiority of the proposed range history and subswath-subband processing has been verified under wide beamwidths, large fractional bandwidths, and wide subswaths through simulation and measured data.

Two-dimensional phase unwrapping is a crucial step in interferometric signal processing. The accuracy of phase unwrapping will significantly affect subsequent processing. Existing algorithms perform poorly in areas with rapid terrain changes or with low coherence. In order to solve the above problems, this letter proposes a method of using neural network to predict residuals, followed by the minimum L ¹ -norm algorithm for phase reconstruction. The network for predicting residuals is called RPNet, and its training phase employs a continuous relaxation strategy, allowing for the construction of more complex, specialized loss terms. The proposed method overcomes the effects of noise and terrain changes due to the accurate prediction of residuals. The compensated phase gradients enable the minimum L ¹ -norm algorithm to reconstruct phases more accurately. The experimental results performed on simulation and InSAR data demonstrate the effectiveness of the proposed method.

When a synthetic aperture sonar (SAS) system operates under low-frequency broadband conditions, the azimuth range coupling of the point target reference spectrum (PTRS) is severe, and the high-resolution imaging range is limited. To solve the above issue, we first convert multi-receivers’ signal into the equivalent monostatic signal and then divide the equivalent monostatic signal into range subblocks and the range frequency subbands within each range subblock in order. The azimuth range coupling terms are converted into linear terms based on piece-wise linear approximation (PLA), and the phase error of the PTRS within each subband is less than π/4. Then, we use the chirp-z transform (CZT) to correct range cell migration (RCM) to obtain low-resolution results for different subbands. After RCM correction, the subbands’ signals are coherently summed in the range frequency domain to obtain a high-resolution image. Finally, different subblocks are concatenated in the range time domain to obtain the final result of the whole swath. The processing of different subblocks and different subbands can be implemented in parallel. Computer simulation experiments and field data have verified the superiority of the proposed method over existing methods.

The convolutional neural network (CNN) has achieved good performance in object classification due to its inherent translation equivariance, but its scale equivariance is poor. A Scale-Aware Network (SA Net) with scale equivariance is proposed to estimate the scale during classification. The SA Net only learns samples of one scale in the training stage; in the testing stage, the unknown-scale testing samples are up-sampled and down-sampled, and a group of image copies with different scales are generated to form the image pyramid. The up-sampling adopts interpolation, and the down-sampling adopts interpolation combined with wavelet transform to avoid spectrum aliasing. The generated test samples with different scales are sent to the Siamese network with weight sharing for inferencing. According to the position of the maximum value of the classification-score matrix, the testing samples can be classified and the scale can be estimated simultaneously. The results on the MNIST and FMNIST datasets show that the SA Net has better performance than the existing methods. When the scale is larger than 4, the SA Net has higher classification accuracy than other methods. In the scale-estimation experiment, the SA Net can achieve low relative RMSE on any scale. The SA Net has potential for effective use in remote sensing, optical image recognition and medical diagnosis in cytohistology.